Expert systems represents a branch of artificial intelligence that deals with computer program solutions to problems normally solved by human experts. Expert systems (and computer-aided medical decision-making) is gaining acceptance in medicine. These techniques may help physicians deal with both the explosion of published medical literature and the increased complexity of decision making (decision analysis, multivariate statistical prediction rules, etc.). We propose to build a Stroke Expert System over 3 years. This system will be designed to run on high-end microcomputers in a clinical setting. We have broken stroke care into component problems: anatomical localization, mechanism of stroke, testing, therapy, and prognosis. The system is conceived as an integrated group of modules (each module dealing with one aspect of the stroke care problem). In addition, a group of supportive modules are planned including an explanatory module, a bibliographic reference module, and a report generator. These modules are integrated by a system executive which communicates with the end user through a uniform interface. Information is exchanged between the component modules through a shared data table (blackboard). Extensive development and validation experiments are planned. Release of an operational expert system is projected for the third year.